Range-separated density-functional theory with random phase approximation applied to noncovalent intermolecular interactions.

Range-separated methods combining a short-range density functional with long-range random phase approximations (RPAs) with or without exchange response kernel are tested on rare-gas dimers and the S22 benchmark set of weakly interacting complexes of Jurecka et al. [Phys. Chem. Chem. Phys. 8, 1985 (2006)]. The methods are also compared to full-range RPA approaches. Both range separation and inclusion of the Hartree-Fock exchange kernel largely improve the accuracy of intermolecular interaction energies. The best results are obtained with the method called RSH+RPAx, which yields interaction energies for the S22 set with an estimated mean absolute error of about 0.5-0.6 kcal/mol, corresponding to a mean absolute percentage error of about 7%-9% depending on the reference interaction energies used. In particular, the RSH+RPAx method is found to be overall more accurate than the range-separated method based on long-range second-order Moller-Plesset (MP2) perturbation theory (RSH+MP2).